565

Biochem. J. (1991) 278, 565-571 (Printed in Great Britain)

Asymmetrical distribution of G-proteins among the apical and basolateral membranes of rat enterocytes Nina

VAN DEN

BERGHE,* Nella J. NIEUWKOOP, Arie B. VAANDRAGER and Hugo R. DE JONGE

Department of Biochemistry

I,

Erasmus University Rotterdam, P.O. Box 1738, 3000 DR Rotterdam, The Netherlands

The distribution of the a and , subunits of guanosine-nucleotide-binding proteins (G-proteins) among the apical and basolateral membranes of polarized rat enterocytes was investigated by ADP-ribosylation assays in vitro and immunoblotting with G-protein-subunit-specific antisera. The enterocytes were found to express ai2, ai3, as and , subunits, whereas ai1 and ao subunits could not be detected. The ac2 and a;3 subunits were located predominantly in the basolateral membrane, in contrast with the a. and ,8 subunits, which were distributed uniformly among both membranes. Furthermore, 39 kDa and 78 kDa proteins, recognized by anti-ail/2 but not anti-ail or anti-a,3 specific antisera, and resistant to ADP-ribosylation by pertussis toxin, were localized exclusively at the apical border. These Gi-related proteins might represent novel members of the G-protein family. Activation of apical G-proteins by GTP or its analogues failed to release the as, ai and /3 subunits or the 39 kDa and 78 kDa a,-like proteins from the membrane, suggesting a functional role for these proteins in the apical membrane itself. Our recent finding of a guanosine 5'-[y-thio]triphosphate-sensitive C1- conductance in the apical membrane of rat enterocytes suggests that one or more of these G-proteins may act as local regulators of specific apical transport functions.

INTRODUCTION

Guanine-nucleotide-binding regulatory proteins (G-proteins) are a family of receptor-coupled signal-transducing proteins that regulate a variety of second messenger systems and ion channels. G-proteins are composed of three subunits, designated a, ,3 and y, with molecular masses of approx. 39-52 kDa, 35-36 kDa and 8-1 1 kDa respectively. The a subunit of each G-protein is unique, whereas the /3 and y subunits appear to be common. The a subunit contains the functional domains for guanine nucleotide binding, GTPase activity, ADP-ribosylation by cholera toxin and/or pertussis toxin, and interaction with specific receptor and effector proteins [1]. For some G-proteins multiple isoforms have been described, caused by alternative splicing of the same gene (e.g. Gs has four isotypes [2]) or by the existence of multiple genes, encoding for similar G-proteins, which form a subfamily (e.g. Gil, Gi2 and G13) [3]. The biological importance of this diversity is as yet not fully understood, since all splice variants of G can activate both adenylate cyclase and Ca2+ channels [2], and all G, species can activate the G-protein gated atrial K+ channel [4]. However, the inhibition of adenylate cyclase in NH108-15 cells was specifically transduced by G12 [5]. It has been generally accepted that G-proteins function mainly at cell surface membranes, where they can associate with various transmembrane receptors and effector proteins. Therefore it is very likely that they are distributed similarly to the receptor/ effector systems to which they are specifically coupled. In enterocytes, as in other epithelial cell types, the plasma membrane is composed of at least two structurally and functionally distinct regions: apical or brush border membranes (BBM) and basolateral membranes (BLM). Most hormone receptors and adenylate cyclase are distributed asymmetrically and are found to be localized predominantly at the BLM [6,7]. In contrast, the BBM of rat enterocytes was found to be enriched in components of the cyclic GMP signalling system [8,9], a G-protein-activated

phosphoinositide cycle [7] and guanosine 5'-[y-thio]triphosphate (GTP[S])-sensitive Cl- channels [10,11]. These findings prompted us to explore the possibility that at least some G-proteins might have an apical localization and might regulate brush-borderspecific functions. Our results indicate that some G-protein subunits (a., /3) are distributed equally among the BBM and the BLM, whereas others are segregated predominantly to the BLM (ai2, ac3), or are localized exclusively in the BBM (39 kDa and 78 kDa a,-like proteins). The potential biological implications of these findings are discussed. MATERIALS AND METHODS Materials Antiserum D43 (kindly provided by L. van der Voorn, Netherlands Cancer Institute, Amsterdam, The Netherlands) was raised in rabbits against a mixture of holomeric G-proteins purified from human brain, and contained antibodies specifically recognizing the /3 subunits and the a subunit of G. on Western blots. The following antisera were kindly donated by Dr. G. Milligan, University of Glasgow (specificities and sensitivities are summarized in [12,13]). (1) Antiserum IMI was raised in rabbits against the peptide NLKEDGISAAKDVK, corresponding to amino acids 22-35 of the a subunit of G., and specifically recognizes a.. (2) Antiserum SG1, produced in rabbits against a decapeptide (KENLKDCGLF) corresponding to the C-terminal amino acid sequence of the a subunit of transducin, specifically recognizes the ailI and the ac2 subunits, but not the acx3 subunit. (3) Antiserum 13B was produced in rabbits against a peptide KNNLKECGLY corresponding to the C-terminal decapeptide of the ac subunit of the inhibitory G-protein Gi3, and specifically recognizes ai3 but not ailI or ac2. (4) Antiserum I1C, specific for the ail subunit and not recognizing the ai2 and or ac3 subunits,

Abbreviations used: G-protein, guanosine-nucleotide-binding protein; G8 and G1, stimulatory and inhibitory G-proteins respectively; GTP[S], guanosine 5'-[y-thio]triphosphate; p[NH]ppG, guanosine 5'-[fty-imido]triphosphate; GDP[S], guanosine 5'-[fl-thio]diphosphate; BLM, basolateral membranes; BBM, brush border membranes; BBMV, BBM vesicles; DTT, dithiothreitol; TBS, Tris-buffered saline. * To whom correspondence should be sent.

Vol. 278

566 was raised against the synthetic peptide LDRIAQPNYI, corresponding to amino acids 159-168 of the a subunit of the

inhibitory G-protein, Gi1. (5) Antiserum CS1, produced in rabbits against a decapeptide (RMHLRQYELL) corresponding to the C-terminal amino acid sequence of the a subunit of the stimulatory G-protein, specifically recognizes as subunits. Antiserum AS/7 (obtained from Du Pont), similar to antiserum SG1, was raised in rabbits against the decapeptide KENLKDCGLF and specifically recognizes the a subunits of transducin, Gil and Gi2. [32P]NAD+ (specific radioactivity 30 Ci/mmol) was obtained from New England Nuclear, Frankfurt, Germany; pertussis toxin was from List Biological Laboratories, Campbell, CA, U.S.A.; cholera toxin was obtained from Sigma; 14C-labelled Rainbow protein molecular mass markers were from Amersham.

Preparation of BLM The small intestine of a rat was removed under light diethyl ether anaesthesia and rinsed twice with 20 ml of ice-cold saline (0.9 % NaCl). All further steps were performed at 0-4 'C. Insideout segments of rat jejunum were mounted on metal rods attached to a vibration apparatus (Vibro-Mixer El from Chemap A.G., Hannedorf ZN, Switzerland). Epithelial cells were released by mechanical vibration (50 Hz, amplitude 1.5 mm) for 30 min in 0.13 M-NaCl/5 mM-EDTA/ 10 mM-Tris, pH 7.4, and isolated by low-speed centrifugation at 345 g. The cells were washed twice in 20 ml of 300 mM-mannitol/12 mM-Tris/HCl, pH 7.4. BLM were prepared essentially as described by Murer et al. [6]. Briefly, cells were homogenized in a Virtis blender, and nuclei, cell debris and mitochondria were removed by differential centrifugation. The remaining membranes were pelleted and BLM were separated from BBM by sucrose gradient centrifugation. The BLM were recovered between 30 and 40 % sucrose, pelleted and finally resuspended in 100 mM-mannitol/10 mM-Hepes/Tris, pH 7.4. The membranes were freshly used or were frozen in liquid nitrogen and stored at -80 'C.

Preparation of BBM: brush border caps and brush border membrane vesicles Intact brush border (BB) caps from small intestinal epithelial cells were isolated by vibration of inside-out gut segments in hypo-osmotic EDTA (2.5 mM) as described by de Jonge [14] and finally resuspended in 20 mM-Tris/HCl, pH 7.3. One type of BBM vesicles (BBMV), hereafter referred to as KSCN-BBMV, were prepared by exposing the brush-border caps to a chaotropic agent (0.52 M-KSCN) as described in detail by Hopfer et al. [15]. Vesicles prepared by this method are virtually depleted of cytoskeletal proteins and calmodutin [15,16]. Another type of BBMV, hereafter referred to as Mg-BBMV, was prepared from isolated enterocytes by mechanical homogenization and differential Mg2+ precipitation as described [17]. The BBM do not aggregate in the presence of divalent cations [18], in contrast with all other subcellular membranes.

ADP-ribosylation of membranes Cholera toxin and pertussis toxin were activated by incubation for 20 min at 37 'C in 4 mM-dithiothreitol (DTT)/ 10 mM-sodium phosphate buffer, pH 8. Conditions for toxin-catalysed ADPribosylation were optimized for each toxin as recommended by Riberio-Neto et al. [19]. The pertussis-toxin-stimulated [32P]ADP-ribosylation assay mixture contained final concentrations of the following reagents in 100 ,ul: 25 mM-Tris/HCl, pH 8, 5 1tM-[32P]NAD' (5 ,uCi/tube), 20 mM-/3-glycerophosphate, 2.5 mM-MgSO4, 0.1 mM-GTP, 1 mM-ATP, 10 mM-creatine phosphate, S ,ug of creatine phosphokinase (25 units/mg), 1 mmEGTA, 2 mM-DTT, 10 mM-thymidine, 0.5 mM-NADP+, the pro-

N. van den Berghe and others tease inhibitors leupeptin (5 ,ug/ml), elastatinol (5 ,ug/ml), chymostatin (2 ,ug/ml) and trypsin inhibitor (20 sg/ml), 5 mMEDTA, 500 ng of preactivated pertussis toxin and 150 ,g of

membrane protein. The cholera-toxin-stimulated [32P]ADPribosylation assay mixture contained final concentrations of the following reagents in 100,ul: 25 mM-sodium phosphate buffer, pH 7,5 ,eM-[32P]NAD+ (5 #uCi/tube), 20 mM-/3-glycerophosphate, 2.5 mM-MgSO4, 0.1 mM-GTP, 1 mM-ATP, 10 mM-creatine phosphate, 5 ,ug of creatine phosphokinase (25 units/mg), 1 mMEGTA, 2 mM-DTT, 10 mM-thymidine, 0.5 mM-NADP+, a mixture of protease inhibitors as above, 10 /M-GTP[S], 5 ,g of preactivated cholera toxin and 150 ,g of membrane protein. Thymidine was added to inhibit poly(ADP-ribosyl)ation and NADP+ was to inhibit NAD+-glycohydrolase activity. When indicated, saponin (final concentration 0.1 %) or alamethicin (25 #g) was also included in the assay. The reaction was carried out at 30 °C for 15 min and stopped by addition of trichloroacetic acid (10 % final concn.). After 10 min on ice, the samples were centrifuged (10 min, 10000 g). The pellet was washed twice with ice-cold double-distilled water and finally solubilized by boiling for 3 min in sample buffer (0.1 M-Tris, 1.7 % SDS, 6.7 % glycerol, 0.7 M-fl-mercaptoethanol, 0.1 % Bromophenol Blue, pH 7) prior to SDS/PAGE.

SDS/PAGE, immunoblotting and autoradiography SDS/PAGE was performed according to Laemmli [20]. Samples for electrophoresis were dissolved in sample buffer, boiled for 3 min and analysed on 10 % polyacrylamide slab gels. Gels were fixed and stained by Coomassie Brilliant Blue, dried and exposed for autoradiography using Kodak SB 5 films. For immunoblotting, proteins were separated on gel and electroblotted on to nitrocellulose paper (0.1 ,um; Schleicher and Schuell) in 25 mM-Tris/192 mM-glycine/20 % methanol. After blotting, non-specific binding was blocked in 3 % gelatin in Trisbuffered saline (TBS: 20 mM-Tris, 0.5 M-NaCl, pH 7.5). The blots were washed twice for 5 min in TTBS (0.05 % Tween in TBS), followed by incubation with antisera (dilutions as indicated elsewhere) for 60 min at room temperature. After three washes with TTBS (10 min/wash), the blots were incubated with 1251I Protein A (0.2 ,uCi/ml of TTBS) for 90 min at room temperature, or with alkaline-phosphatase-conjugated goat anti-rabbit antibodies for 60 min at room temperature. Blots were then washed for 5 x 10 min, air-dried and exposed for autoradiography using Kodak SB 5 films, or colour was developed using 0.8 mg of 5bromo-4-chloro-3-indolyl phosphate and 1.65 mg of NitroBlue Tetrazolium in 5 ml of 0.2 M-Tris/10 mM-MgCl2, pH 9.1. The reaction was stopped in water.

G-protein subunit release experiments Open BB caps were isolated, resuspended in 100 mmmannitol/lO mM-Hepes/Tris, pH 7.4, and 200-300 ,tg samples were incubated for various time periods at 30 °C in the presence or absence of 100 /tM of GTP[S], guanosine 5'-[/Jy-imido]triphosphate (p[NH]ppG) or guanosine 5'-[fi-thio]diphosphate (GDP[S]) and 0.5 mM-MgC12. The incubation was stopped by ultracentrifugation for 10 min at 207 kPa in a Beckman Airfuge at room temperature. The proteins in the supernatant were precipitated with 10 % trichloroacetic acid, centrifuged (10 min at 10000 g, 0 °C), and solubilized in 100 ,u1 of SDS sample buffer (approx. 10% of total protein was recovered in this fraction). The pelleted particulate fraction was directly solubilized in 100 ,ul of SDS sample buffer (approx. 90 % of total protein). A volume of 10 ,1 of each fraction was analysed by SDS/PAGE and immunoblotting. In some experiments, protease inhibitors were added during the incubation (SO,uig of trypsin inhibitor/ml, 5 mM-phenylmethanesulphonyl fluoride and 20 ,g of leupeptin/ 1991

567

Asymmetrical distribution of G-proteins in enterocytes ml). In other experiments the membranes were first ADPribosylated and then separated into particulate and soluble fractions.

1

RESULTS

4

mass | (kD)

(kDa) 69-

Molecu lar Imass (kDa)

J48

46-

46 -42

30- ...

Isolation of the BBM and BLM of enterocytes Three distinct preparations of apical membranes were used in this study: (1) non-vesiculated BB caps, consisting of apical membrane, cytoskeletal proteins and the terminal web adjacent to the microvilli, (2) KSCN-BBMV, obtained by removing

Table 1. Distribution of marker enzymes among BLM and BBM fractions of rat enterocytes

Enzyme activities are expressed relative to those in the homogenate of isolated rat small intestinal epithelial cells. Sucrase is used as a marker of the BBM and Na+/K+-ATPase as a marker of the BLM. Values are mean + S.D. from three experiments. Relative specific activity Fraction

Homogenate BLM BB caps Mg-BBMV KSCN-BBMV

1

Sucrase

Na+/K+-ATPase

1.0 0.2+0.1 8.5 +0.5 6.7+0.8 17.6+0.4

1.0 15.0+2.0 2.0+0.5 1.3 +0.8 1.2+0.3

2 * :.Y.:.

mass (kDa) 94 67-

3

4

,. ,

'A

Molecular

I

,

i

-

-

43-

-

30-t

4 41 kDa

-

qx si,"

Fig. 1. Pertussis-toxin-dependent ADP-ribosylation of intestinal epithelial membrane proteins Autoradiographic profiles are shown of 3iP-labelled proteins (10 jag/lane) after incubation of BLM (lane 1), Mg-BBMV (lane 2), BB caps (lane 3) and KSCN-BBMV (lane 4) with [32PINAD+, preactivated pertussis toxin and 0.1 % saponin. Experimental details of the ADP-ribosylation assay are as in the Materials and methods section. Migration of molecular mass standards (kDa) was as indicated. The ADP-ribosylation of the 41 kDa protein was observed only in the presence of pertussis toxin (results not shown). The results are representative for at least three other experiments.

Vol. 278

3

MolecularMoeua mass

Other procedures Protein concentrations were determined according to Lowry et al. (21] using BSA as standard. Sucrase was assayed with sucrose as a substrate by the method of Dahlqvist [22]. Na+/K+ATPase was assayed by its ouabain-sensitive phosphatase activity, according to the method described by Adams et al. [23].

2

Fig. 2. Cholera-toxin-dependent ADP-ribosylation of intestinal epithelial membrane proteins

Autoradiographic profiles are shown of 32P-labelled proteins (10 ,ug/lane) after incubation with ["P]NAD', preactivated choleratoxin and 0.1 % saponin of BLM (lane 1) and Mg-BBMV (lane 2); after incubation with [32P]NAD' and preactivated cholera-toxin of BB caps (lane 3); and after incubation of [32P]NAD', preactivated cholera-toxin and 25 ,ug of alamethicin of KSCN-BBMV (lane 4). Experimental details of the ADP-ribosylation assay are as in the Materials and methods section. Migration of molecular mass standards (kDa) was as indicated. ADP-ribosylation of the 46 and 42 kDa proteins was only observed in the presence of cholera-toxin (results not shown). The results are representative of three or more experiments.

cytoskeletal and peripheral membrane proteins (including the terminal web) from the BB caps by KSCN treatment, and (3) Mg-BBMV, isolated by differential Mg2" precipitation, containing both apical membrane and microvillar cytoskeletal proteins, but not the terminal web. The purity of the membranes was verified by measuring the specific activity of sucrase as a BBM marker [24] and of Na+/K+ ATPase as a marker of BLM [25]. As shown in Table 1, in comparison with the homogenate, all three preparations of BBM were enriched in sucrase (7-17-fold), and the BLM preparation was enriched in Na+/K+-ATPase (1 5-fold), confirming the relative purity of both membrane fractions. As judged by the marker enzyme distribution, the BLM were relatively devoid of BBM contamination and vice versa. Identification of G. subunits by pertussis- and cholera-toxin-catalysed ADP-ribosylation G-proteins can be distinguished by their ability to serve as substrates for pertussis-toxin-catalysed ADP-ribosylation (G1, GO), for cholera-toxin-catalysed ADP-ribosylation (G.), for both pertussis- and cholera-toxin-catalysed ADP-ribosylation (transducin) or by a lack of recognition sites for either of the toxins. These properties were exploited to identify G-protein species in the enterocyte. Incubating BB caps, KSCN-BBMV, Mg-BBMV and BLM with [32P]NAD+ and pertussis toxin revealed the presence of a 41 kDa protein in all membrane fractions (Fig. 1). This 41 kDa protein was tentatively identified as the a subunit of G1, since the other pertussis toxin substrate, transducin, is uniquely present in rods and cones, and a specific antiserum against ao failed to detect such subunits in these membrane fractions (see Fig. 3d). Cholera-toxin-catalysed ADPribosylation enabled the detection of two other proteins with molecular masses of 46 kDa and 42 kDa, tentatively identified as splice variants of the a subunits of G, (Fig. 2) [2], which were likewise present in all membrane fractions. These subunits sometimes appeared as doublets, possibly as a result of covalent modifications.

N. van den Berghe and others

568 (a)

(b) 1

2

3

4

Molecular mass (kDa) 200 > 92.5 . 69 .

Molecuilar mass (kIDa) 200 > 9 2.5

:.:.

i

...

1

2

3

4

:..:.. ;::

*.:-

C

*::

....

-78 kDa

.. :: : :.

:.

69 . 46 >

46 - 41 kDa

30>

41 kDa -39 kDa

30P Dye .

Dye >

(d) MolecLular mass (klDa) 200

(c) 2

1

3

4

Molecular mass (kDa)

1

2

3

4

9,2.5

200 > 69

92.5k 69 46.: 46 *

___0

-*

,.

a

-46 kDa -42 kDa

30o

-

30h Dye

-36 kDa

Dye o

Fig. 3. Detection of G-protein subunits with subunit-specific antisera in intestinal epithelial membrane fractions Portions of 5 jug of BLM (lane 1), Mg-BBMV (lane 2), BB caps (lane 3) and KSCN-BBMV (lane 4) proteins were separated by SDS/PAGE. Immunoblotting was performed as described in the Materials and methods section. (a) Antiserum 13B, az3-specific, dilution 1: 200; (b) antiserum SG1, ail/2-specific, dilution 1:200; (c) antiserum CSl, a.-specific, dilution 1:200; (d) antiserum D43, ao- and 4-specific, dilution 1:O000. Migration of the molecular mass standards (kDa) was as indicated. The results shown are representative of at least three other experiments.

(b)

(a) 2

1 -41 kDa

g,

(c) 1 2

2 kDa 39 kDa

-41 -

-46 kDa -42 kDa

Fig. 4. Distribution of ADP-ribosylated substrates in particulate and soluble fractions of BB caps BB caps (150 ,ug) were ADP-ribosylated by pertussis toxin or cholera toxin as described in the Materials and methods section without addition of detergent, and the soluble and particular fractions were prepared. Both fractions were solubilized in 70 jul of SDS sample buffer and analysed on SDS/PAGE. Following electrophoresis, the ADP-ribosylated proteins were transferred to nitrocellulose paper and identified by autoradiography. (a) Particulate (lane 1) and soluble (lane 2) fractions of pertussis-toxin-catalysed ADPribosylated BB caps. (b) Treatment of the same blot with antiserum SG1 (specific for ail/2; dilution 1:200) following decay of 32P radioactivity. Detection was done with 125I-Protein A and autoradiography. (c) Particulate (lane 1) and soluble (lane 2) fractions of cholera-toxin-catalysed ADP-ribosylated BB caps. The results shown are representative of three experiments.

Identification of G-protein subunits by specific antisera A variety of specific antisera against a and , subunits were used to further identify the different G-proteins. Antiserum 13B,

specific for the ai3 subunit, identified a protein of 41 kDa predominantly in the BLM, although a faint staining could be detected in the BBM fractions (Fig. 3a). Antiserum SG1, which specifically recognizes both ail and aci2 subunits, also showed a substantial enrichment of a 41 kDa protein in the BLM compared with the BBM (Fig. 3b). Since antiserum I1C, specific for the a, 1 subunit, did not react with a 41 kDa protein in any of the membrane fractions (results not shown), the protein recognized by antiserum SG1 is tentatively identified as the ai2 subunit. As indicated by combined [32P]NAD+ labelling and immunoblotting, the pertussis-toxin-sensitive 41 kDa protein shown in Fig. 1 comigrated with the proteins identified by antisera SGl (Figs. 4a and 4b) and 1 3B (results not shown) and with purified G-proteins from human and bovine brain (see Figs. 5 and 6), further confirming their identity as ai subunits. Antiserum SG1 additionally recognized a 39 kDa protein, which was detected only in the BB caps and occasionally in KSCN-BBMV, but not in the Mg-BBMV or the BLM (Fig. 3b), and which was clearly distinguishable from the 41 kDa band (Figs. 4b and 5). In contrast with the 41 kDa ai subunit, this 39 kDa protein was not a substrate for pertussis-toxin-catalysed ADP-ribosylation (Figs. 1 and 4). This antiserum also detected a 78 kDa protein predominantly present in Mg-BBMV, although it was additionally found in BB caps and BLM, but not in KSCN-BBMV (Fig. 3b). The staining of the 41, 39 and 78 kDa proteins was blocked by the peptide to which the antiserum was raised, arguing against 1991

:*.

569

Asymmetrical distribution of G-proteins in enterocytes 1

2

3

4

5

6

7

8

9 10 11 '..

As s,m&

Molecular mass (kDa) .4 78

441

Fig. 5. Release of the 39 kDa ace-like protein from BB caps Suspensions of 300 ,ug of BB caps were incubated for 0, 5, 15 and 30 min at 30 °C and then separated into particulate (lanes 2, 4, 6, 8 and 10 respectively) and soluble (lanes 3, 5, 7, 9, and 11, respectively) fractions. In lanes 10 and 11, the incubation was carried out for 30 min at 30 'C in the presence of protease inhibitors (see the Materials and methods section). Particulate and soluble fractions 100 of SDS sample buffer and analysed on were solubilized in l1 SDS gels. Proteins were detected with antiserum SG1 (aiIl/2-specific, dilution 1: 200), followed by alkaline phosphatase-conjugated second antibody. As a marker, a purified G-protein mixture from bovine brain was loaded (lane 1, 0.2 jug). The results are representative of at least three other experiments.

(a) 1 2

Molecular mass (kDa)

78i

3

(b) 1

.:.:: .

2

...

3

.:

:.:

...

:.

.:

..............

:. ::::: .::

41

.

39>

_

..........

.:: .:

Fig. 6. Peptide displacement of immunoreactive bands in BB caps recognized by a specific antiserum against ccl/2 BB caps (300 fig) were incubated for 5 min at 30 'C and then separated into particulate and soluble fractions (lanes 2 and 3 lO of SDS respectively). Both fractions were solubilized in 100 sample buffer and analysed on gels. Lane 1 contains 0.2 ,g of a purified G-protein mixture from bovine brain. Proteins were detected with antiserum SG1 (ail/2-specific; dilution 1:200), followed by alkaline phosphatase-conjugated second antibody. (a) SG1; (b) SG1 presaturated with the peptide to which it was raised (0.5 mg/ml, 5 min incubation at 20 'C) and subsequently diluted 1: 200.

non-specific binding (Fig. 6). Similar results were obtained with a different antiserum (AS/7) raised against the same conjugate (results not shown). Antiserum CS1, specific for as subunits, identified two proteins of 46 kDa and 42 kDa (which sometimes appeared as doublets) in both the BLM and the BBM fractions (Fig. 3c). These 46 kDa and 42 kDa proteins co-migrated with the ADP-ribosylated cholera toxin substrates shown in Fig. 2, and are likely to represent two species of a; formed by alternative splicing [2]. The membrane rather than cytoskeletal origin of a; was indicated by its relative abundance in cytoskeleton-free KSCN-BBMV in comparison with Mg-BBMV and BB caps (Fig. 3c). The distribution of the a; subunit among the BBM and the BLM was virtually symmetrical (Fig. 3c, lanes 1 and 4). Antiserum D43, specific for ao and , subunits, identified only a 36 kDa band, which co-migrated with the /3 subunit, in all Vol. 278

membrane fractions (Fig. 3d), albeit slightly enriched in the BBM. This shows that the /? subunit is symmetrically distributed in enterocytes. The ao subunit could not be detected in any of these membrane fractions either by antiserum D43 (Fig. 3d) or by a second antiserum, IM 1 (results not shown). In contrast, both antisera were capable of detecting ao subunits in crude membrane fractions of brain (D43; results not shown) or of NG108-15 cells (IMI, [13]). G-protein subunit release experiments To examine whether apical G-proteins function at the apical membrane or are released and translocated to other subcellular regions, release studies were performed in vitro using the nonvesiculated BB caps as a G-protein source. First we investigated whether pertussis and cholera-toxin-catalysed ADP-ribosylation could trigger release of a, and a. respectively from the open BB caps. As shown in Fig. 4(a), the ADP-ribosylated ai subunit remained confined to the 100000 g particulate fraction and was not translocated to the supernatant. This was confirmed by immunoblotting of the 32P-labelled BB caps with specific antisera against ac 1/2 (Fig. 4b) and ai3 (results not shown). Also, incubation of the BB caps with GTP[S] or p[NH]ppG for 60 min at 30 °C did not trigger ai release, either alone or in combination with ADP-ribosylation (results not shown). The ADP-ribosylated as subunits were predominantly recovered in the particulate fraction, although some as subunits were detected in the soluble fraction (Fig. 4c). Immunoblotting with specific antiserum against a. confirmed these data and further showed that this release occurred also in the absence of cholera toxin and was therefore not triggered by the ADP-ribosylation assay (results not shown). Incubation ofthe BB caps with GTP[S] or p[NH]ppG for 60 min at 30 °C did not enhance the release (alone or in combination with ADP-ribosylation), nor did GDP[S] inhibit basal release (results not shown), suggesting that the small basal release of a. was aspecific and not induced by activation of G,. In contrast with the ;i and as subunits, the 39 kDa a,-like protein recognized by antiserum SG1 was readily released after incubation of the membranes at 30 'C. This release was independent of guanine nucleotides and could not be inhibited by GDP[S]. Solubilization of the protein was already detectable after 5 min of incubation, and increased with time (Fig. 5, lanes 2-9). The addition of protease inhibitors prevented solubilization (Fig. 5, lanes 10 and 11). This suggests that the release of the 39 kDa protein is triggered by endogenous proteases that become activated/solubilized during incubation of the BB caps at higher temperature. As solubilization of the 39 kDa protein was not accompanied by a measurable shift in its molecular size (Fig. 5), the release mechanism is likely to involve the proteolysis of another protein to which the 39 kDa protein is coupled. The proportion of 39 kDa protein releasable in 15-30 min varied between experiments and is most likely dependent on the protease content of the isolated membranes, which can vary per BB batch and may include both intrinsic proteases and contaminating pancreatic enzymes. In some experiments, complete release was found in 15-30 min (Fig. 4b). Immunoblotting did not reveal solubilization of the /J subunit (results not shown), confirming the hydrophobic nature of this protein. Addition of 0.1 % BSA as a carrier protein before precipitation with trichloroacetic acid did not influence the results (not shown). DISCUSSION

Membrane-associated G-proteins generally act as signal transducers, coupling receptors for hormones, neurotransmitters or light to effectors, such as adenylate cyclase, phospholipases, ion

570 channels and phosphodiesterases [1]. Considering the asymmetrical distribution of most, if not all, G-protein-coupled hormone receptors at the basolateral pole of the enterocyte [7], the G-proteins are likewise expected to reside preferentially in this subcellular region. In our immunochemical experiments, such a distribution pattern could be substantiated for some of the G-proteins, notably the ax2 and ai3 subunits, but not for others: the as and the fi subunits were present in about equal concentrations in both membranes. Two newly detected ai-like proteins (39 and 78 kDa), identified with antiserum raised against ail/2, were found almost exclusively in the apical border. ai and a. subunits could be detected in pertussis- and choleratoxin-catalysed ADP-ribosylation assays respectively. With respect to the ai subunits, however, the ADP-ribosylation data and the immunoblotting data were qualitatively but not quantitatively comparable. The ADP-ribosylation reaction is known to be critically dependent on assay conditions and addition of detergent. In our experience, the labelling of each membrane preparation changed dramatically when different detergents were used, and their order of potency varied with different membrane species. To ensure optimal assay conditions, different detergents were therefore selected for a comparison between the various membrane preparations examined (Fig. 2). As the ADPribosylation under our assay conditions is presumably incomplete and variable between different preparations, these data are unsuitable for quantitative analysis. Besides the classical G-proteins, G12, G13 and G., two possibly novel Gi-like species were detected immunochemically in the intestinal membranes. The 39 kDa protein, detected by a specific antiserum against a, l /a,2 but not a,il and ac3, differed from the 41 kDa ai subunits in that: (1) its molecular mass was slightly lower; (2) it did not serve as a substrate for pertussis-toxincatalysed ADP-ribosylation; (3) it was found only in BB caps and (at a low level) in KSCN-BBMV, but not in BLM; (4) it was readily released after incubation of BB caps at 30 'C. As the Nterminus of several a subunits is known to be important for fl/y interaction and membrane attachment [26], one could argue that this 39 kDa protein originates from ax2 through limited proteolysis of an N-terminal segment. Such a hypothesis would imply that: (1) proteolysis is limited and restricted to a small Nterminal fragment, because the 39 kDa protein is still recognized by antiserum SGI (raised against the C-terminal decapeptide); (2) the C-terminus of a, (essential for recognition of the a subunit by pertussis toxin [26]) must be altered, causing the lack of pertussis-toxin-catalysed ADP-ribosylation of the 39 kDa protein shown in Figs. I and 4, without however disturbing antiserum recognition; (3) the protease responsible for this cleavage is located exclusively at the apical border; no evidence for a 41 to 39 kDa protein conversion was found in purified BLM; (4) proteolysis must have occured in vivo or during preparation of the BBM; in contrast, several other peripheral membrane proteins known to be sensitive to proteolysis (e.g. cyclic GMPdependent protein kinase; villin) do not undergo proteolytic modification during isolation of BB caps [27]. However, if the 39 kDa protein arises by proteolysis of the 41 kDa aj2 subunit, it remains unclear why prolonged incubation of the brush borders, leading to a time-dependent increase in solubilization of the 39 kDa band, does not significantly diminish the intensity of the 41 kDa band, in clear contrast with the immunoreactive 78 kDa band (Fig. 5). Alternatively, the 39 kDa protein may originate at least in part from proteolysis of the 78 kDa protein, which in some (Fig. 5), although not all, experiments decreased when release of the 39 kDa protein increased. Although the size of the 78 kDa protein suggests that it might be a dimer of the acx2 subunit or of the ai-like 39 kDa protein, no dissociation into 39-41 kDa subunits was observed following N-ethylmaleimide

N. van den Berghe and others treatment or boiling in the presence of ,-mercaptoethanol. Similarly to the 39 kDa protein, the 78 kDa protein is readily released from the membrane and its release could be prevented by protease inhibitors (Fig. 5). Its distribution is different from both ac2 and the a,-like 39 kDa protein in that it is highly enriched in the Mg-BBMV (Fig. 3b). A similar high-molecularmass protein was also detected by the same antiserum in rat liver membranes [28]. The rat liver 78-90 kDa protein was likewise enriched in fractions that contained few or no ai subunits [28]. The function of this protein is not yet known, but its molecular mass argues against a role as a classical heterotrimeric G-protein. Assuming that the 39 kDa protein is a novel member of the cx1 family, its ability to be readily released from the BB by proteolysis makes it difficult to draw conclusions about its intracellular localization. It remains possible that this protein is more uniformly distributed among BLM and BBM in the intact cell, but is lost during the preparation of the BLM and the Mg-BBMV, but not of the BB caps. The possibility is rather remote since MgBBMV, in contrast with open BB caps, are formed immediately after shearing the cells and are resealed tightly in a right-side-out orientation, preventing the release of intramicrovillar proteins [17]. A more likely possibility is that this protein is associated with the terminal web structure which is recovered only in BB caps, but not in Mg-BBMV or cytoskeleton-depleted KSCNBBMV. Such a localization might have important implications for its physiological function. The protein could be involved in endo- or exo-cytotic pathways or in the regulation of the organization of the cytoskeleton, as described for G-proteins in neutrophils [29]. Assuming, however, that the 39 kDa protein is a proteolytic fragment of the 78 kDa protein, the unequal distribution of these proteins among Mg-BBMV (enriched in 78 kDa protein) and BB caps (enriched in 39 kDa protein) could reflect the protection of intravesicular proteins in Mg-BBMV against degradation by extravesicular proteases, preventing a possible conversion of the 78 kDa protein into the 39 kDa protein (cf. [17,27]). Clearly, further studies are needed to purify and characterize both the 78 kDa and the 39 kDa ai-like proteins in order to identify their possible relationship and their physiological significance. The targets for the apical G-proteins do not need to be in the BBM itself. It has already been observed that G-protein a subunits can be released from the membrane upon activation by GTP, GTP analogues or ,-adrenergic receptor activation [30-32]. This is of particular interest in understanding the mechanism of action of cholera toxin. Enterocytes are the target for this toxin in vivo, causing secretory diarrhoea. It has always been intriguing how cholera toxin, entering the enterocyte from the luminal site, is able to activate adenylate cyclase, which is located at the BLM. Dominguez et al. [33], who were the first to demonstrate apical localization of a.s subunits in rabbit enterocytes by the use of non-immunological techniques, have suggested that cholera toxin activates ax. in the BBM, which is then translocated and couples to adenylate cyclase in the BLM. However, in the present study in vitro we could not detect release of a., ai or /1 subunits from the BBM, even after prolonged exposure to GTP or its analogues or after cholera or pertussis-toxin-catalysed ADP-ribosylation. Under similar conditions, a slow and partial release of a. subunits has been described in rat glioma membranes [31]. The reason for this tissue difference is unclear. Although the absence of Gprotein subunit translocation needs to be confirmed for intact epithelium, our data support the concept that the apical Gproteins fulfil a functional role in the membrane to which they are targeted. In this regard, at least part of the cellular content of G, and G. is expected to be localized in the BLM, given their known interaction with adenylate cyclase in this tissue. Our study shows that this is indeed true for rat enterocytes. 1991

Asymmetrical distribution of G-proteins in enterocytes So far, no physiological activators or inhibitors of apical Gproteins in epithelial cells have been defined. Possible candidates include unidentified receptor proteins in the luminal membrane, second messengers activated by hormone receptors in the BLM or, by analogy with Go in the growth cone [34], a GAP-43-like protein, which can intracellularly activate G-proteins without the necessity for a G-protein-coupled receptor. A polarized distribution of G-protein subunits was also found in the rat renal cell line LLC-PK1 by Ercolani et al. [35]. In the renal epithelial cells, in contrast with the enterocyte, a;3 was localized exclusively in the BBM, where it is possibly involved in the activation of amiloride-sensitive Na+ channels [36] and apical Cl- channels [37]. A similar role for G-proteins in the apical membrane of intestinal epithelial cells has been suggested by our recent identification of a novel type of Cl- channel in this membrane which was activated by GTP and GTP analogues and inhibited by GDP[S] in the absence of cytoplasmic messengers [10,1 1]. The presence of a G-protein-activated phosphatidylinositol cycle in the BBM [7] also suggests a role for apically localized G-proteins. The a. subunits and the 39 kDa/78 kDa a,-like proteins are possible candidates for both Cl- channel and phospholipase C activation, in contrast with the xi subunits, which are localized almost exclusively in the BLM. We thank Dr. G. Milligan for generously supplying the specific antisera to ai1l, ai2, a13, ao and a.. We are also grateful to L. van der Voorn for her gift of the specific antiserum to a/fl and the G-protein mixture purified from human and bovine brains. This study was supported by the Netherlands Organization for the Advancement of Scientific Research (NWO).

REFERENCES 1. Gilman, A. G. (1987) Annu. Rev. Biochem. 56, 615-649 2. Mattera, R., Graziano, M. P., Yatani, A., Zhou, Z., Graf, R., Codina, J., Birnbaumer, L., Gilman, A. G. & Brown, A. M. (1989) Science 243, 804-807 3. Jones, D. T. & Reed, R. R. (1987) J. Biol. Chem. 262, 14241-14249 4. Yatani, A., Mattera, R., Codina, J., Graf, R., Okabe, K., Padrell, E., Iyengar, R., Brown, A. M. & Birnbaumer, L. (1988) Nature (London) 336, 680-682 5. McKenzie, F. R. & Milligan, G. (1990) Biochem. J. 267, 391-398 6. Murer, H., Ammann, E., Biber, J. & Hopfer, U. (1976) Biochim. Biophys. Acta 433, 509-519 7. Vaandrager, A. B., Ploemacher, M. C. & de Jonge, H. R. (1990) Am. J. Physiol. 259, G410-G419 8. de Jonge, H. R. (1975) FEBS Lett. 53, 237-242 Received 26 September 1990/27 February 1991; accepted 23 April 1991

Vol. 278

571 9. de Jonge, H. R. (1981) Adv. Cyclic. Nucleotide Res. 14, 315-333 10. de Jonge, H. R., van den Berghe, N., Tilly, B. C., Kansen, M. & Bijman, J. (1989) Biochem. Soc. Trans. 17, 816-818 11. Tilly, B. C., Kansen, M., van Gageldonk, P. G. M., van den Berghe, N., Galjaard, H., Bijman, J. & de Jonge, H. R. (1991) J. Biol. Chem. 266, 2036-2040 12. Green, A., Johnson, J. L. & Milligan, G. (1990) J.. Biol. Chem. 265, 5206-5210 13. Mullaney, I., Magee, A. I., Unson, C. G. & Milligan, G. (1988) Biochem. J. 256, 649-656 14. de Jonge, H. R. (1981) Cold Spring Harbor Conf. Cell Proliferation Protein Phosphorylation 8, 1313-1332 15. Hopfer, U., Crowe, T. D. & Tandler, B. (1983) Anal. Biochem. 131, 447-452 16. Vaandrager, A. B., Ploemacher, M. C. & de Jonge, H. R. (1986) Biochim. Biophys. Acta 856, 325-336 17. van Dommelen, F. S., Hamer, C. M. & de Jonge, H. R. (1986) Biochem. J. 236, 771-778 18. Louvard, D., Maroux, S., Baratti, J., Desnuelle, P. & Mutaftschiev, S. (1973) Biochim. Biophys. Acta 291, 747-763 19. Ribiero-Neto, F. A. D., Matteca, R., Hildebrandt, J. P., Codina, J., Field, J. B., Birnbaumer, L. & Sekura, R. D. (1985) Methods Enzymol. 109, 566-572 20. Laemmli, U. K. (1970) Nature (London) 227, 680-685 21. Lowry, 0. H., Rosebrough, N. J., Farr, A. L. & Randall, R. J. (1951) J. Biol. Chem. 193, 265-275 22. Dahlqvist, A. (1968) Anal. Biochem. 22, 99-107 23. Adams, R. J., Cohen, D. W., Gupte, S., Johnson, J. D., Wallick, E. T., Wang, T. & Schwartz, A. (1979) J. Biol. Chem. 254, 1240412410 24. Kenny, A. J. & Maroux, S. (1982) Physiol. Rev. 62, 91-128 25. Fujita, M., Ohta, H., Kawai, K., Matsui, H. & Nakao, M. (1972) Biochim. Biophys. Acta 274, 336-347 26. Osawa, S., Dhanasekaran, N., Woon, C. W. & Johnson, G. L. (1990) Cell 63, 697-706 27. de Jonge, H. R., Schmeeda, H. & Shaltiel, S. (1987) Eur. J. Biochem. 169, 503-509 28. Ali, N., Milligan, G. & Evans, W. H. (1989) Mol. Cell. Biochem. 91, 75-84 29. Therrien, S. & Naccache, P. H. (1989) J. Cell Biol. 109, 1125-1132 30. McArdle, H., Mullaney, I., Magee, A., Unson, C. & Milligan, G. (1988) Biochem. Biophys. Res. Commun. 152, 243-251 31. Milligan, G. & Unson, C. G. (1989) Biochem. J. 260, 837-841 32. Ransnas, L. A., Svoboda, P., Jasper, T. R. & Insel, P. A. (1989) Proc. Natl. Acad. Sci. U.S.A. 86, 7900-7903 33. Dominguez, P., Velasco, G., Baros, F. & Lazo, P. S. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 6965-6969 34. Strittmatter, S. M., Valenzuela, D., Kennedy, T. E., Neer, E. J. & Fishman, M. C. (1990) Nature (London) 344, 836-841 35. Ercolani, L., Stow, J. L., Boyle, J. F., Holtzman, E. J., Lin, H., Grove, J. R. & Ausiello, D. A. (1990) Proc. Natl. Acad. Sci. U.S.A. 87, 4635-4639 36. Light, D. B., Ausiello, D. A. & Stanton, B. A. (1989) J. Clin. Invest. 84, 352-356 37. Schwiebert, E. M., Light, D. B., Eejes-Toth, G., Naray-Eejes-Toth, A. & Stanton, B. A. (1990) J. Biol. Chem. 265, 7725-7729